Statistical conundrums when designing adaptive trials in oncology

1st May 2018 (Last Updated August 9th, 2019 12:14)

Arena International’s Clinical Operations in Oncology Trials West Coast 2018 wrapped up last week, and the roundtable exploring adaptive trial designs in oncology saw a focus on the sometimes overlooked role of focused statistician engagement when designing these studies.

It appears companies, particular smaller firms with low resources, are not engaging the right statistical expertise early on. Specialist biostatisticians are often coming on board too late, when the need arises to rectify and resurrect failed trials.

Delegates on the roundtable included those from small pharma, big pharma, specialty contract research organisations (CRO), big CROs, trial vendors and consulting firms, and questions rather than answers fueled the discussion during the conference, which took place in Burlingame, California.

While the term “adaptive design” gets thrown around a lot and often is a buzz word, there are different ways to design an adaptive trial. This might involve ways to restore the size of the trial, changing dose selection, dropping arms being tested in the trial, tweaking endpoints and amending the hypothesis or the trial’s patient population, explained Natasa Rajicic, senior director, Cytel, a biostatistics consulting firm based in Cambridge, Massachusetts. What cannot change is messing around with the analysis plan, she noted.

Despite it being ideal for statisticians to come in at the start of a program - even at the hypothesis stage - it is not the typical practice, noted Rajicic. Instead, sponsors often outsource statistical responsibilities to CROs in a pay-for-service model that covers a more basic and general statistical review of the patient sample size, she said. It is common for sponsors to bring in the biostatistician consultation at the Phase III planning stage, she noted, adding many companies come to Cytel for consulting advice when they are too far progressed.

A biostatisticial perspective

Lyne Cantin, clinical research manager, Stiris Research, a specialized oncology CRO based in Canada, said her firm has learnt the importance of leveraging biostatisticians early on in the dialogue to develop clinical trial programs. The standard way to approach trial statistics is to engage a statistician at the end of the trial protocol design process to calculate the number of patients that should be recruited, she said, noting biostatisticians have the potential to offer much more.

Biostatisticians bring a different perspective that could potentially enhance the study design, and they could be particularly helpful in informing the trial hypothesis if they understand the disease and patient response variabilities, Cantin said. Keeping statisticians distant from protocol and hypothesis discussions and confining them to running the numbers needs to change. In fact, involving them earlier might actually impact their evaluation on the number of patients that should be recruited, she said.

Rajicic has prior experience at Pfizer, where the pharma giant had a dedicated statistics team involved in early trial design dialogue with clinical operations and pharmacology teams. However, she said, this was a relatively unique structure and culture. Smaller companies also do not have the luxury of such resources, she noted.

Navigating FDA approvals

A number of CRO and small biotech delegates are dealing with navigating the US Federal Drugs Authority (FDA), which is often rigid on approvals. They are also trying to avoid penalties when tweaking protocol designs mid-trial, which raises the possibility of needing to increase sample sizes.

The FDA stated on 5 February - at a workshop held at Duke University - that it is willing to accept clear prespecified plans with trial changes noted upfront, Rajicic said. The most important consideration is to have a prespecified plan with the regulator and decide why a trial adaptation is needed and what it will try to achieve by changing the path mid-trial, delegates noted during the roundtable.

In terms of ultimate goals for the approved label and having those at the forefront of discussions when designing trials, delegates noted the need for a pre-emptive strategy that may involve changing labels after interim analyses, when patient populations need to be modified. At the end of the day, this is a lot easier said than done and will require a comprehensive and earlier dialogue between sponsor, clinicians, trial coordinators and statisticians.

by Surani Fernando

Surani Fernando is the US Editor of GlobalData’s investigative journalism team, which is focused on breaking exclusive news and providing nuanced, forward-looking analyses.. To access more investigative news like this article, visit GlobalData.